FactPEGASUS: Factuality-Aware Pre-training and Fine-tuning for Abstractive Summarization

TL;DR

FactPEGASUS is a new abstractive summarization model that enhances factual accuracy through improved pre-training and fine-tuning techniques, outperforming existing models in factuality on multiple tasks.

Contribution

It introduces a fact-aware pre-training strategy and three novel fine-tuning components to improve the factuality of abstractive summaries.

Findings

Significantly improves factuality metrics and human evaluations.

More robustly maintains factuality in zero-shot and few-shot settings.

Does not rely solely on extractiveness for factual accuracy.

Abstract

We present FactPEGASUS, an abstractive summarization model that addresses the problem of factuality during pre-training and fine-tuning: (1) We augment the sentence selection strategy of PEGASUS's (Zhang et al., 2020) pre-training objective to create pseudo-summaries that are both important and factual; (2) We introduce three complementary components for fine-tuning. The corrector removes hallucinations present in the reference summary, the contrastor uses contrastive learning to better differentiate nonfactual summaries from factual ones, and the connector bridges the gap between the pre-training and fine-tuning for better transfer of knowledge. Experiments on three downstream tasks demonstrate that FactPEGASUS substantially improves factuality evaluated by multiple automatic metrics and humans. Our thorough analysis suggests that FactPEGASUS is more factual than using the original pre-training objective in zero-shot and few-shot settings, retains factual behavior more robustly than strong baselines, and does not rely entirely on becoming more extractive to improve factuality. Our code and data are publicly available at: https://github.com/meetdavidwan/factpegasus